Spatial Nutrient Models Calibrated by Field Residual Mapping
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital nutrient models for agricultural fields are not agronomically field-specific, leading to overestimation or underestimation of nutrient requirements due to unaccounted field characteristics, and require extensive data that is often unavailable, making them inefficient and unreliable.
Innovation Solution
An agricultural intelligence computer system that calibrates a nutrient model to a specific field using historical data, computing residual values and generating model correction data by comparing these values to spatial mappings of field characteristics, thereby accounting for unique field properties.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a generalized predictive model is used that treats each field as a sum of inputs, then the model can be computed efficiently with available data, but the model cannot account for unique field characteristics leading to overestimation or underestimation of nutrient requirements
Solution Approach 1:
The patent segments the nutrient modeling problem into two parts: a generalized predictive model that handles common field characteristics efficiently, and a residual correction model that accounts for unique field-specific factors. This segmentation allows the system to maintain computational efficiency while improving accuracy by addressing field-specific deviations separately.
Solution Approach 2:
The patent introduces residual values as correction parameters that adjust the output of the generalized model. By computing residuals between actual and predicted nutrient requirements, the system dynamically changes the model parameters to account for unmeasured field characteristics, thereby improving accuracy without requiring complete re-modeling.
2Measurement precision
If an individual nutrient model is created for a specific field using massive amounts of data, then the model can be highly accurate for that field, but such data is rarely available making the approach unfeasible and inefficient
Solution Approach 1:
Instead of requiring complete field-specific data, the patent uses partial information in the form of residual values computed from limited available data. The residual correction approach applies only the necessary adjustment rather than attempting to build a complete individualized model, making the system feasible with available data while still improving accuracy.
Solution Approach 2:
The patent introduces residual values as an intermediary between the generalized model and field-specific requirements. These residuals act as a mediator that captures the effect of unmeasured field characteristics without requiring direct measurement of all field-specific parameters, thereby reducing data collection complexity while maintaining model accuracy.
3Adaptability or versatility
If a digital fertility model is used without calibration to field-specific characteristics, then the model can be applied broadly across multiple fields, but the model becomes biased due to unmeasured, poorly measured, or unknown field properties
Solution Approach 1:
The patent implements a feedback mechanism where residual values computed from actual field performance are used to correct the generalized model's predictions. This feedback loop allows the model to adapt to field-specific characteristics while maintaining its broad applicability, thereby improving reliability without sacrificing versatility.
Solution Approach 2:
The patent dynamically changes model parameters by introducing residual corrections that adjust the generalized model's output for each field. This parameter adjustment allows the same base model to be applied across multiple fields while accounting for field-specific variations, maintaining both versatility and reliability.
Data Source
AI summary
In an embodiment, an agricultural intelligence computing system stores a digital model of crop growth, the digital model of crop growth being configured to compute nutrient requirements in soil to produce particular yield values based, at least in part, on data unique to an agricultural field. The system receives agronomic field data for a particular agronomic field, the agronomic field data comprising one or more input parameters for each of a plurality of locations on the agronomic field, nutrient application values for each of the plurality of locations, and measured yield values for each of the plurality of locations. The system computes, for each location of the plurality of locations, a required nutrient value indicating a required amount of nutrient to produce the measured yield values. The system identifies a subset of the plurality of locations where the computed required nutrient value is greater than the nutrient application value. The system computes, for each of the subset of the plurality of locations, a residual value comprising a difference between the required nutrient value and the nutrient application value. The system generates a residual map comprising the residual values at the subset of the plurality of locations. Using the residual map and the one or more input parameters for each of the plurality of locations, the system generates and stores particular model correction data for the particular agronomic field.